fix: Try improving relay_datagram_send_channel()

[matheus23]

Description

Closes #3067

Since multiple Connections can have multiple AsyncUdpSocket IpPollers, it's incorrect to assume that a single AtomicWaker can wake up these tasks correctly.
Instead, if there's only one AtomicWaker for all of them, only the last registered waker will be woken when there's capacity again.

I'm changing this such that poll_writable will actually add the current task's waker to a list, if it hasn't been added yet.
And successfully recving an item will wake all tasks.

A second issue was that between the call to self.sender.capacity() (it being 0) and self.waker.register, another thread might have successfully recvd an item and called self.waker.wake().
This means that the first thread could've registered a waker, while the capacity is actually non-zero. Not terrible (it's very likely that it'll be woken up when the capacity jumps to 2 this time), but also not great.
The fix is to re-check the capacity after registering the waker.
The downside is that we keep the waker and thus potentially get a spurious wakeup, but that's fine.

Notes & open questions

This... "fixes" the concerns, but I'd actually like to write a loom test for the concerns eventually.

It's just... a lot of work, and I'd rather prioritize other things, but also improve this part of the code at the same time.

Change checklist

• Self-review.
• Documentation updates following the style guide, if relevant.
• Tests if relevant.
• All breaking changes documented.

[github-actions]

Documentation for this PR has been generated and is available at: https://n0-computer.github.io/iroh/pr/3118/docs/iroh/

Last updated: 2025-01-10T13:41:04Z

[github-actions]

Netsim report & logs for this PR have been generated and is available at: LOGS
This report will remain available for 3 days.

Last updated for commit: 7db1d47

[matheus23]

Netsim results:

test	case	throughput_gbps	throughput_transfer
iroh	1_to_1	1.41	1.41
iroh	1_to_3	4.26	4.27
iroh	1_to_5	7.20	7.21
iroh	1_to_10	12.25	12.27
iroh	2_to_2	3.03	3.03
iroh	2_to_4	5.57	5.57
iroh	2_to_6	9.03	9.05
iroh	2_to_10	13.69	13.74

versus main:

test	case	throughput_gbps	throughput_transfer
iroh	1_to_1	1.38	1.38
iroh	1_to_3	4.27	4.28
iroh	1_to_5	6.99	7.00
iroh	1_to_10	12.01	12.03
iroh	2_to_2	3.06	3.07
iroh	2_to_4	5.86	5.87
iroh	2_to_6	9.20	9.22
iroh	2_to_10	13.62	13.65

No noticable difference.

No differences in relay-only performance (cargo run --release -p iroh-bench --all-features -- iroh --only-relay).

[flub]

nice!